Market barriers to energy efficiency

Richard B. Howarth and Bo Andersson

Discussions of energy policy in an environmentally constrained world often focus on

the use of tax instruments to internalize the external efSects of energy utilization or
achieve spectfied reductions in energy use in the most cost-effective manner. A
substantial literature suggests, however, that signtficant opportunities exist to reduce
energy utilization by implementing technologies that are cost-eflective under
prevailing economic conditions but that are not fully implemented by existing market
institutions. This paper examines the theory of the marketfor energy-using equipment,
showing that problems of imperfect information and transaction costs may bias
rational consumers to purchase devices that use more energy than those that would
be selected by a well-informed social planner guided by the criterion of economic
eflciency. Consumers must base their purchase decisions on observed prices and
expectations of post-purchase equipment performance. If it is dtficult or costly for
individuals to form accurate and precise expectations, the level of energy ejiciency
achieved by competitive markets will vary from the socially e#icient outcome. Such
‘market barriers’ suggest a role for regulatory intervention to improve market
performance at prevailing energy prices.
Keyword.s: Energy efficiency; Market barriers: Imperfect information

The efficacy of the market mechanism in achieving a device exceeded the cost of the improvement,
socially desirable level of energy efficiency has competitive producers could increase their profits by
generated considerable debate in policy circles and the exploiting consumers’ willingness to pay for the
academic literature. According to the simplest change. It follows under these assumptions that all
neoclassical models of competitive markets, perfectly efficiency measures that were cost-effective at prevailing
informed, rational individuals consider the total costs prices would be realized by competitive markets. The
of their decisions when purchasing energy-using problem of energy policy would thus be reduced to
equipment. If the value of the energy savings ensuring that energy prices reflected the full social
generated by improving the energy efficiency of a costs of energy production and utilization (Sutherland
C251).
The validity of this idealized model as a description
Richard B. Howarth is with the Environmental Studies
of observed reality has been cast into doubt by
Board, University ofcaiifornia, Santa Cruz, CA 95064, USA
microeconomic studies of the costs and benefits of
and Bo Andersson is with the Stockholm School
energy-efficient technologies. Hausman [lo], for
of Economics, Stockholm, Sweden.
example, conducted a study of consumer purchases of
The authors acknowledge support from the Norwegian Research air conditioners and found that consumers could
Council, the Swedish National Energy Administration, and the
Stockholm School of Economics. Additional funding was provided
achieve considerable present value savings by
by the Stockholm Environment Institute and the US Environmental switching from the models actually purchased to more
Protection Agency under US Department of Energy Contract No energy-efficient alternatives. Observed consumer
DE-AC03376SF00098. Lam Bergman, Olav Bjerkholt, Kjell Arne
Brekke, Eystein Gjelsvik, Jon Koomey, Jim McMahon, Richard
behaviour would be consistent with lifecycle cost
Norgaard, Lee Schipper, Steinar Strom, and especially Alan Sanstad minimization only at a discount rate of 25% per year
provided valuable comments and suggestions. averaged across income groups. Implicit discount
Final manuscript received 13 April 1993. rates fell sharply with higher incomes; low-income

262 0140/9883/93/040262~11 0 1993 Butterworth-Heinemann Ltd

 Market barriers to energy e@ciency: R. B. Howarth and B. Anderson

groups behaved as though they used discount rates of full social benefits of research and development
up to 89%. In a similar vein, Ruderman et al [20], expenditures, especially when markets are fragmented
focusing on sales of residential appliances and heating or patent protection is easily circumvented through
and cooling equipment, estimated implicit discount ‘copycat’ inventions (Arrow [l]; Kamien and
rates ranging from 20 to 800% per year - well above Schwartz [ 111). Such eventualities suggest a potential
the returns available on other investments. role for government in supporting research and
The factors behind the so-called ‘efficiency gap’ ~ development (Fisher and Rothkopf [S]). Indeed, the
the differential between the level of energy-efficiency empirical literature suggests that existing government-
actually achieved and the level judged to be sponsored projects to develop energy-efficient tech-
cost-effective at prevailing prices ~ have been explained nologies have reaped rich rewards (Geiier et al L6J;
mainly in terms of ‘market barriers’ related to Rosenfeld [19]).
consumer decision-making (Carlsmith et al [2]). The existence of an efficiency gap suggests the
Consumers are said to lack full information regarding possibility that energy use may be reduced relative to
available energy-efficient technologies, use a discount status quo levels at a net cost saving. A study by the
rate that is ‘too high’ in evaluating investments in US National Academy of Sciences [28], for example,
energy efficiency, or simply lack the means to carefully concluded that energy-related emissions of carbon
evaluate the consequences of their decisions. In some dioxide could be reduced by as much as 37% through
cases, such as the rental housing market, the costs and the adoption of energy-efficient technologies that are
benefits of energy efficiency improvements may accrue cost-effective under current economic conditions.
to different agents, splitting the incentive to invest in As Grubb [9] frames the issue, the achievement of
otherwise attractive technologies. Finally, transactions an ideal level of energy efficiency involves a two-step
costs may swamp the often small net benefits process (Figure 1). First, energy prices should be set
achievable by individual consumers (Krause and Eto to reflect the full social costs of energy production
C121). and utilization, including external environmental
Issues relating to the supply of energy-using costs. Proper pricing shifts the energy supply curve
equipment may also induce barriers to the development inward from the S to S*. Second, policies should be
and adoption of energy-efficient technologies. It is well implemented to improve market performance given
known that private firms are unable to capture the prevailing prices, shifting the demand curve from the

Price of Energy

I Pricing Policy

AS

‘--------D
Removal
of Market
\ I Barrie;s
Y D*

E* E Energy Use

Figure 1. Two-stage process to achieve energy efficiency.

ENERGY ECONOMICS October 1993 263

Market barriers to energy eficiency: R. B. Howarth and B. Andersson

laissez faire D to the cost-effective D* without produce as ordinarily inefficient cars of comparable
reductions in energy services. Only when these size’. This claim is apparently based on anecdotal
measures are implemented in tandem is energy use accounts of recent developments in vehicle technology;
reduced from the status quo E to the socially efficient in the absence of careful data on manufacturing costs
level E* without incurring offsetting social costs.’ and vehicle performance, its general validity is difficult
Such arguments have had a significant influence on to ascertain.
energy policy. In some jurisdictions, governments Second, heterogeneity in the population of energy
require utilities to exploit opportunities to reduce users may imply that technologies that are cost-
electricity demand as part of their long-term planning. effective on average are not cost-effective for some
Environmentalists are arguing for tougher automobile classes of consumers. For example, although electric
fuel economy standards on the grounds that such heating systems are generally more costly than natural
standards would provide not only energy but also gas systems on a lifecycle cost basis, electric space
economic savings. Appliance efficiency standards are heaters may be cost-effective if used infrequently (as
now an accepted part of US energy policy and are in vacation homes) since their capital requirements
under consideration in other nations. are low. Considerable attention to detail is required
The economic underpinnings of the presumed to identify the true costs of improved energy efficiency.
efficiency gap, however, have not been firmly This paper presents a theoretical analysis of
established. While a plausible case may be made that demand-side market imperfections that drive a wedge
market barriers impede the adoption of certain between the cost-effective level of energy efficiency and
energy-efficient technologies, little effort has focused the level that is realized by the interaction of rational
on the analysis of these barriers using formal models agents in the marketplace. We find that consumer
rooted in economic theory. The argument that perceptions regarding the performance of energy-using
consumers use especially high discount rates in equipment have an important influence on the market
evaluating the present value costs of energy-using potential of energy-efficient devices. Even when
technologies, for example, is theoretically objectionable. consumers correctly estimate and seek to minimize the
Why would a rational person pass up a 20&800% total costs of equipment ownership, anomalies in the
return when alternative investments yield only transmission of information between producers and
5-lo%? Either the facts are wrong, consumers are consumers may result in an inadequate level of energy
‘irrational’ in the sense that they do not evaluate efficiency in unfettered competitive markets.
energy-using technologies in a manner consistent with The range of market barriers to energy efficiency
lifecycle cost criteria, or there is more to the story considered here is by no means exhaustive. The paper’s
than meets the eye. purpose, however, is not to construct a comprehensive
Before proceeding further, we wish to establish two theory but rather to shed some light on an interesting
caveats. First, there is some disagreement regarding and vexing set of issues. A principal finding is that
the potential for cost-effective energy savings. In-depth frictionless models of competitive equilibrium are an
engineering studies, for example, indicate that raising incomplete and potentially misleading guide to energy
the fuel economy of new US automobiles beyond policy. Good policy arguably involves more than
36-44 miles per gallon of gasoline by the year 2000 simply ‘getting prices right’. A potential role exists for
(as compared to the current level of 28 miles per gallon governments to intercede when the vagaries of market
in government tests) would entail a tradeoff between institutions lead to lags in the development and
higher costs and reductions in vehicle amenities such adoption of energy-efficient technologies.
as size and safety (Difiglio et al [3]; Ledbetter and
Ross [13]). Lovins and Lovins ([14], p 477), on
Energy efficiency and consumer bebaviour
the other hand, maintain that prototype vehicles
averaging some 71 to 92 miles per gallon present While there is considerable variation in the range of
opportunities for cost-effective energy savings because energy-using technologies, raising the efficiency of a
such vehicles ‘should cost about the same to mass device generally increases the cost of manufacturing it
but reduces the downstream costs borne by users.
Equipment manufacturers thus have no direct
‘This point was made in the mid-1970s by Schipper (1231, pp
incentive to produce efficient devices and will do so
457-58) who argued that ‘energy conservation [is] the strategy of
adjusting and optimizing energy-using systems and procedures so only to the extent demanded by consumers. We show
as to reduce energy requirements while holding constant or below that competitive markets will fail to generate a
reducing total [sy%em]-costs ,. The largest stimulus to more socially efficient level of energy efficiency if their
efficient energy utilization will occur in response to direct economic
incentives and governmental policies designed to aid those structural characteristics impede the frictionless
incentives.’ transfer of information between market participants.

264 ENERGY ECONOMICS October 1993

 Market barriers to energy ejiciency: R. B. Howarth and B. Andersson

Of course the market will not converge on a socially income groups imply discount rates substantially in
efficient outcome unless energy prices reflect the full excess of the market rate of interest. The gap between
social costs of energy production and utilization. By the discount rates of low-income consumers and the
‘socially efficient’, we mean that the costs and benefits returns available on capital markets suggests that
of energy utilization - and investments in improved significant social benefits may be garnered through
equipment performance - are equated at the margin. programmes providing loan guarantees or direct
In our analysis we take prices as given and investigate subsidies to improve the energy efficiency of low-
the response of market participants to prevailing income households.
market conditions.
As we noted above, several alleged imperfections
A one-period model
have been identified in the market for energy-using
equipment. One of the first that comes to mind is To furtherexnj~re _-some
. .._ of
-_ the iwws
____ _____- raked
_---_- nhnve
-__ ._ WP
.._
consumers’ possible lack of information about the construct formal models that illustrate how competitive
characteristics of available technologies. One can markets will fail to generate a socially efficient level
safely assume that there is generally a positive cost of energy efficiency if their structural characteristics
associated with the dissemination of information. impede the effective transmission of information
Equipment performance may be difficult to observe, between market participants. Consider the following
and the costs of improving consumer information may one-period model. There is a large number of identical
exceed the expected benefits to private agents. individuals who divide their incomes between
Principal-agent problems may also impede the expenditures on an energy-using good (‘a device’) and
adoption of energy-efficient technologies. In the other consumption items. Each device uses e units of
typical example from the rental housing market, energy while the price of energy is 4. The total cost of
landlords would bear the costs of investments in purchasing and using a device is thus p+qe.
improved energy efficiency while the resulting energy Consumers are fully cognizant of their incomes and
savings would accrue to the renters who pay the utility relative prices but are unable to observe directly the
bills.’ If tenants were unable to observe perfectly energy intensity of devices prior to sale. Based on
building characteristics and transaction costs exceeded advertising, discussions with sales representatives, and
the benefits achievable through such investments, other information, however, consumers form a point
cost-effective energy savings would be passed up even expectation e* regarding equipment performance.
in a world of rational agents (Fisher and Rothkopf Taking income and other prices as given, the demand
cv ). for devices is thus x(p+qe*) with x’(.)<O. Given a
Consumers’ uncertainty about the benefits of choice between devices with differing energy intensities,
improved energy efficiency may also impede the consumers would choose the model that minimized
uptake of energy-efficient technologies. As Sutherland the ex ante ownership cost p+qe*.
[25] notes, rational consumers may act as though they On the producer side, a large number of atomistic
are using high implicit discount rates when making firms produce devices at the unit cost c(e) such that
decisions under a high degree of uncertainty. c’(e) < 0 and c”(e) >O. Thus costs are convex and
Subjective uncertainty, however, may stem from the increase monotonically as energy efficiency improves
fact that precise estimates of energy prices and or energy intensity falls. Suppose that we take income
equipment performance are costly to obtain from the and energy prices as given and assume that the device
perspective of individual consumers. If the costs of market is in competitive equilibrium. What are the
gathering information were pooled across individuals, relevant equilibrium conditions pertaining to the price
substantial economies of scale could be achieved which and energy intensity of devices? Clearly the price must
could reduce the uncertainties associated with certain be set equal to the incremental cost of manufacturing
technologies. a device, or else new firms could enter the market at
Finally, a number of authors have noted that lower prices and earn positive profits. Thus p=c(e),
low-income households lack the access to credit and profits are zero.
required to finance investments in improved energy This, however, is only part of the solution since
efficiency. Hausman’s [IO] findings on this matter are costs depend on energy intensity and e is an
highly suggestive: the appliance choices of low- endogenous variable. If consumers held perfect
:.-P- --,_I’
miormdtion about device energy intensity, we should
‘Landlords would be unwilling to pay for utilities since tenants expect an equilibrium to minimize the total cost of
would then have an incentive to overutilize energy services. Tenants
owning and using a device. Were this not the case, an
would be unwilling to make investments in improved energy
efficiency if they expected to move out before the benefits of the opportunity would exist to produce a device better
investments were substantially realized. suited to consumer preferences, and producers would

ENERGY ECONOMICS October 1993 265

Market hurriers to energy .$ciency: R. B. Howarth and B. Andersson

presumably rush to fill the void. In formal terms, an efficient level of energy intensity may arise even when
equilibrium with perfect information must minimize consumers use biased estimators of device performance.
c(e) + qe through the choice of e, a problem that yields Note, however, that k#O implies an inefficient level
the first-order condition of device ownership since consumers will misjudge the
total ownership costs on which their purchase
c’(e)= -q (1) decisions are based.
Conversely, conformance between consumers’
The interpretation of this condition is straightforward. equilibrium estimates of energy intensity and actual
At the margin, the cost of reducing the energy intensity device performance is insufficient to ensure social
of a device is set equal to the market value of the efficiency. Suppose, for example, that subjective
energy savings achievable through the change. Higher expectations are given by f(e) = (e + 1)/2, the unit
energy prices justify higher production costs, so energy cost function is c(e) = l/e, and the price of energy is
intensity is a decreasing function of energy prices. q = 2. In equilibrium, these assumptions imply that
Provided that the price of energy reflects its full social consumer expectations and the prevailing level of
costs and that producers are using the best available energy intensity are identical at e* = e = (2/q)‘12 = 1
technology, it is clear that this outcome is socially as compared to the socially efficient level of (1/2)1’2.
efficient since it equates the marginal costs and benefits Thus it is not the level of expectations per se that is
of energy-intensity reductions. In a world of perfect relevant in determining market behaviour. Instead, it
information, consumers signal their preferences to is the relationship between expectations formation and
producers through their market behaviour, and the true characteristics of energy-using equipment,
producers respond by producing goods that satisfy expressed here by f(e).
those preferences as fully as possible. Why would consumers base their expectations on
Note, however, that the model does not impose the anything but the observed characteristics of energy-
restriction that consumers hold perfect information using equipment? In reality, it may be costly or difficult
regarding device energy intensity. Because they must to obtain precise and accurate estimates of equipment
commit to their purchase decisions based on performance, justifying the use of simple rules of
potentially imperfect expectations about equipment thumb when making purchase decisions. While a firm
performance (Russell and Thaler [21]), consumers will might advertise its product as achieving a particular
favour devices that minimize the ex ante ownership performance standard, such information might be
cost p+qe* whether or not the resulting outcome is discounted by consumers who base their beliefs on
optimal ex posf. Suppose that the expected and actual factors such as their personal experience with
levels of energy intensity are related by the equipment ownership.
differentiable function e* =f(e). This function captures Suppose, for example, that an individual consumer
the psychological processes that shape a consumer’s could observe a device’s true energy intensity for a
beliefs based on the objective information at her/his fixed cost C. Would it be rational for her to incur this
disposal. Then the level of energy intensity that cost? Assume for simplicity that the individual’s device
prevails in competitive equilibrium must solve the demand was fixed at x and that a firm was willing to
problem supply an efficient device at the price PeSJ = c(eef/)
such that c’(eeff)= -q such that ex post ownership
Minimize (p+qe* = c(e) + qf(e)} costs were minimized. Then the net ex post benefit of
(2)
obtaining perfect information would be

The first-order condition for this problem takes the (P + qe - P,~/ - qe,ss)x - C (4)
form
This expression could be either positive or negative
c’(e) = -qf’(e) (3) depending on the parameters of the model. Where
information costs were high, however, a rational
This expression differs significantly from the efficiency consumer would have a clear incentive to base her
condition defined by Equation (l), and it is clear that behaviour on imperfect expectations of equipment
the two will coincide only where f’(e) = 1. In performance.
particular, the function f(e) = e + k will satisfy this The analysis presented here indicates that the
criterion for any constant k, not just the obvious case formation and content of consumer beliefs regarding
where k=O so that there is agreement between the performance of energy-using equipment are
consumer expectations and the realized outcome. This fundamental to the market potential for energy-
is rather surprising, because it implies that a socially efficient devices. Because this subject has been

266 ENERGY ECONOMICS October 1993

 Market barriers to energy eficiency: R. B. Howarth and B. Andersson

investigated by behavioural scientists, we need not of firms is large so that individual firms are unable
base our opinions on pure speculation or theoretical to affect the prevailing ex ante ownership-cost
assertions. In summarizing the literature, Stern level, differentiation of Equation (5) and algebraic
concludes that the information held by consumers manipulation yield the first-order conditions
regarding residential energy use ‘is not only
incomplete, but systematically incorrect. Generally Pit = 4%) (6)
speaking, people tend to overestimate the amounts of
energy used by - and that may be saved in ~ 6( 1 - a)c’(e,)x, - c’(ei, _ 1)xit _ 1- ct6qxi, = 0 (7)
technologies that are visible and that must be activated
each time they are used’ ([24], p 205). Such findings Equation (6) holds that each firm sets the cost of its
support the general theme of our argument and call devices equal to its marginal cost. This condition,
attention to the role of expectations in understanding which is necessary for efficient resource allocation, is
consumer behaviour. familiar from the static model. Because marginal cost
is equal to average unit cost, each firm’s profits are
zero.
Dynamics with adaptive expectations
The equilibrium condition expressed by Equation
The relationship between consumer expectations and (7) suggests potentially complicated dynamics; a full
equipment performance is clarified by the following analysis of the possibilities is beyond the scope of this
dynamic version of the model. There are n identical discussion. Note, however, that the socially efficient
firms that supply devices to the market. The ith firm level of energy intensity (e,rf) is found by equating
sells xit units at price pir and energy intensity ei, at the marginal costs and benefits of intensity improve-
sequential dates t=O, 1, . . The unit cost of ments so that c’(eess) = -4. Since Equation (7) does
producing a device is c(e,,). As above, costs are convex not conform to this rule, we should not expect
and decreasing in energy intensity so that c( .) < 0 and competitive markets to achieve an efficient level of
c”(. ) > 0. The price of energy is constant at q, and energy intensity given the structure of the model.
devices are used only during the period in which they The analysis of steady-state equilibria yields further
are purchased. The ex post cost of owning a device insights into the relationship between consumer
bought from firm i is thus pit + qe,,. expectations and realized equipment performance. In
Consumers purchase devices separately at each date steady state, the level of energy intensity is constant
in full cognizance of their incomes and prevailing over time and identical between firms at e, and
prices. They cannot observe the true energy intensity Equation (7) reduces to
of devices prior to sale. Based on past equipment
performance and manufacturer
consumers form an expectation
reputation, however,
regarding the energy c’(e)= -q(,_y+,,)
intensity of each firm’s devices according to the
recurrence relation ez = (1 - u)e,“;_ I + cte, _ 1. Here
CIE (0, 1) is a constant that captures the rate at which Because ad/( 1 - 6 + ~6) < 1 under the restrictions
expectations are updated. Consumers view devices imposed on the parameters, we may conclude that
from each manufacturer as providing equivalent c’(e)>c’(eess) = -4. Since the unit cost function is
energy services and will purchase a model only if it decreasing and convex with respect to energy intensity,
minimizes expected ownership costs. It follows that this implies that e>eefJ so that the level of energy
the ex ante ownership cost (P,) must be equivalent for intensity prevailing in competitive equilibrium exceeds
each firm in the market so that P, = pit + qe,? for the socially efficient level. While we have not explored
each i. the dynamics of the model under general conditions,
We assume that firms seek to maximize the it may be shown that energy intensity converges
intertemporal profit level monotonically to the steady state under simplifying
assumptions - for example, when device demand is
cost-inelastic and constant over time.
In steady state, the expected and realized values of
energy intensity are in agreement so that e,“;= e, for
= f d’[P, - qes - c(ei,)]xi, each i. Thus the inefficiency associated with market
(5)
equilibrium does not arise from consumer mis-
perceptions per se. Nor does the inefficiency require
where 8~ (0, 1) is a discount factor used to convert long lags in the adjustment of expectations. Although
future revenues to present-value units. If the number energy intensity is, under the assumptions of the

ENERGY ECONOMICS October 1993 267

Market barriers to energy eficiency: R. B. Howarth and B. Anderson

model, a decreasing function of the lag coefficient CI, offer rebates to home buyers who claimed that their
the fact that firms discount future revenues relative to heating bills were higher than advertised unless the
the present implies that energy intensity is ‘too high’ disparity was very large indeed. Finally, the problem
even when CI= 1 so that expectations adjust in a single of ‘moral hazard’ might arise under either energy
period. Instead, the market veers from an efficient service contracts or performance guarantees. To the
outcome because difficulties in observing the perform- extent that the energy intensity of a device depends
ance of equipment sold on today’s market impedes on user behaviour, institutions that weaken user
the efficient transfer of information between producers incentives to minimize direct energy costs might lead
and consumers. to reduced energy efficiency.
While the precise specification of the model is of
course open to debate, the intuition that supports it
The role of uncertainty
is rather appealing. The fact that consumers must
purchase equipment before they incur the resulting The discussion so far assumes that consumers are
energy costs means that in practice expectations identical in their beliefs and hold point expectations
formation is critical to purchase decisions. It is regarding equipment performance. In reality, however,
reasonable to suppose that consumer expectations will the content of a consumer’s beliefs depends on her
rely heavily on manufacturer reputation and previous cognitive processes and on the specific context in
experience in owning and operating equipment ~ which she receives and evaluates information. It is
factors which reflect past rather than present likely, therefore, that beliefs will differ across
equipment performance. In the steady-state equilibria individuals according to some frequency or probability
described above, expected and realized levels of energy distribution. A related issue is that consumers may
intensity are in exact agreement, and consumers have recognize the fallibility of their expectations and take
no incentive to change their mode of expectations this uncertainty into account as they render decisions.
formation. More generally, consumers will seek to Both of these factors have important implications for
improve the quality of the information at their disposal the uptake of energy-efficient technologies that are
only where the expected benefits outweigh the best addressed using stochastic models of market
costs. Thus expectational errors may persist where behaviour.
information is costly to obtain. Suppose there are two types of devices that provide
The point might be raised that equipment suppliers, equivalent energy services. The unit cost and energy
who presumably know both the marginal cost of intensity of each class of equipment are fixed at ci and
reducing energy intensity and the related marginal e,, and the supply industry is competitive so that the
benefits, would have an incentive to offer contracts to device price pi is set equal to unit cost for i= 1, 2. The
provide device services at a cost below the level implied price of energy is q so that the total cost of owning a
by the model. An energy service company, for example, device of type i is pi + qei.
could offer to provide both a device and the energy There are I? consumers who each purchase exactly
required to operate it for a fixed price below prevailing one device of either class 1 or 2. Each consumer knows
ex ante ownership costs. Alternatively, manufacturers the values of pl, p2 and q with certainty. Consumers
might guarantee the energy intensity of their devices. do not, however, know the true values of e, and e2
In the absence of transaction costs, such an but form unbiased estimates drawn from the
arrangement would lead to an equilibrium with a distributions
socially efficient level of energy intensity. Indeed,
private contractors provide such services in certain ei* = ei + ui (9)
equipment markets.
Energy costs, however, are often small both in where ui is independently and normally distributed
absolute terms and as a fraction of the total cost of with mean 0 and variance CF. As in the models
owning and operating equipment. Thus the private presented above, consumer expectations regarding
gains to be won through the direct provision of energy equipment performance are based on imperfect
services may swamp the associated marketing costs observation of the products offered for sale.
for some technologies. Similarly, the costs of contract Suppose that e, > e2 and p1 + qe, >pz + qe,. These
monitoring and enforcement might well exceed assumptions imply that technology 1 is clearly inferior
uncertainties about energy costs, ruling out perform- because both its true energy intensity and ownership
ance guarantees as an effective marketing tool. It is cost exceed those of technology 2. Will technology 2
notoriously difficult to establish the field performance therefore drive techology 1 off the market? To answer
of energy-using equipment (Nadel and Keating [17]). this question requires a hypothesis concerning
In practice, few would expect a building contractor to consumer behaviour.

268 ENERGY ECONOMICS October 1993

 Market barriers to energy eficiency: R. B. Howarth and B. Anderson

Consider first the case where each consumer naively a threshold that increases with their degree of
takes her estimates of e, and e2 as the true values of uncertainty.
these variables and purchases technology 1 if Why is this an interesting case? Generally speaking,
the most energy-efficient technologies are new to the
market, and it is reasonable to suppose that consumers
will lack familiarity with such products and thus attach
If the number of consumers is large and each forms a high degree of uncertainty to their performance. In
her/his energy intensity estimates separately, the a dynamic world, consumer uncertainty would
difference ez -e: will take on large values with presumably narrow with time and experience. In the
non-zero probability so that this condition will be meantime, however, rational consumer behaviour
satisfied for some portion of the population. This mode permits the sustained use of outdated technologies.
of behaviour therefore implies a positive market share As in the case of the deterministic one-period model
for a clearly inferior technology. discussed above, we may consider the ex post value
Now suppose that consumers recognize that their of perfect information for the model in question.
estimates of e, and e2 are stochastic. In particular, we Suppose that consumers may reduce the variance of
shall assume that they know the true values of g’ their energy intensity estimators 0; (i= 1,2) to zero
(i= 1,2) and that they seek to maximize an expected for a fixed cost C. Then the net ex post value of the
utility function defined over their incomes net of device information cannot exceed
and energy expenditures. Assuming that consumers
exhibit constant absolute risk aversion, they will (14)
purchase technology 1 if
If the difference in user costs between the two
E{-expC-~(I-p,-qe,)lle:} technologies were small relative to the cost of
>E{-expC-~.(I-p2-qe2)lIef} (11) obtaining perfect information, a rational consumer
would choose to make her/his purchase decisions
where 2” is a positive parameter and I is consumer under uncertainty rather than incur costs yielding
income. Since E[exp(Ax)] =exp(p+;1’a2/2) if x is ambiguous net benefits.
normally distributed with mean p and variance c?,
this condition may be written in the form Policy implications

3.q2a: The models outlined above suggest several possible

- exp
i
-1.
(I-pp,-qqe:--
2 )1
” 2 2
policy options and strategies. An obvious suggestion
is that the government or some other independent
agency should monitor and certify the performance of
> -exp -3” I_p,_qe~-!qA (12)
2 energy-using equipment. Indeed, this approach has
i ( >)
been put into practice in the USA, where labels are
required on appliances and automobiles that indicate
Algebraic manipulation then shows that consumers their energy intensity. While the approach is most
will favour technology 1 if practical for mass produced items, it is sometimes used
in the housing market, where buildings that meet
;1q2a: iq20: specific criteria may be certified as being energy
p2+qeZ+ -->pl +qeT+- (13)
2 2 efficient (Wilson et al [29]).
On theoretical grounds, one would expect such
This relation is closely related to the naive decision labels to improve the information held by consumers
rule given by Equation (8); the two coincide when to at least some extent. Empirical work has found,
cr: = g:. As above, some consumers will purchase each however, that performance labelling has ambiguous
technology, with individual decision6 depending on effects on consumer decisions (McNeil1 and Wilkie
the particular signals received. In general, the [16]; Robinson [18]). Such a finding need not’ be
uncertainty consumers attach to their expectations inconsistent with the behavioural models set out
alters their perceptions of the relative attractiveness above. The cognitive processes by which consumers
of the two technologies. This is clearly seen in the case assess and internalize information are clearly relevant,
where 0: = 0 so that consumer uncertainty is limited and it may be costly or difficult to convey information
to the more efficient technology. With this assumption, to consumers in a manner that leads them to make
consumers will purchase technology 2 only if its ‘correct’ decisions. As Sanstad et al [22] frame the
expected cost savings relative to technology 1 exceed issue, ‘it is not hard to reach a reductio ad absurdum

ENERGY ECONOMICS October 1993 269

Market barriers to energy efJiciency: R. B. Howarth and B. Andersson

in which the government undertakes the education in McMahon et al [lS] (see also US Department of
basic engineering economics of consumers upon their Energy [27]) estimate that these standards will save
arrival at the department store appliance section’. 24 EJ of primary energy between 1990 and 2015. While
A second approach would be the implementation the standards will increase equipment costs by US832
of taxes or subsidies that adjusted the signals received billion (109) in present-value terms, the associated
by market participants. In our static model of the energy savings are valued at US!%78 billion. On
device market under point expectations, for example, balance, then, the standards are expected to yield net
an efficient allocation of resources would result if the savings of US$46 billion.
government taxed device manufacturers at the unit Of course, the conformance of energy intensity to
rate q[e -f(e)] where f(e) is, as above, a representative a socially efficient level does not mean that consumers
consumer’s expectation of device energy intensity. The correctly anticipate the ex post cost of equipment
equilibrium price of devices would then be ownership, and this suggests that the demand for
devices may fail to converge to an efficient level
P = c(e) + qCe -f(e)1 (15) through the implementation of performance standards
alone. In the models we have considered, this problem
and consumers’ ex ante expectations concerning the could be overcome through the taxation or subsidization
costs of device ownership and ex post social costs of devices to span the gap between expected and
would equilibrate at realized ownership costs. Here it is the content rather
than the functional determination of expectations
P+f(e)=c(e)+qCe+f(e)l+qf’(e)=c(e)+qe (16) which is relevant.
The variability of energy prices between regions
An equilibrium would minimize ex ante ownership raises interesting questions with respect to equipment
costs, implying the first-order conditions performance standards. Should standards be uniform
across regions, or should they be fine-tuned to reflect
c’(e) + q[ 1 -f’(e)] + qJ“(e) = 0 (17) local conditions? Tollison and Ekelund [26] examined
this question for natural gas furnaces in the USA,
so that c’(e) = -4. Thus the social costs and benefits finding that increasing returns to scale in manufacturing
of reducing energy intensity would be brought into favoured the implementation of uniform standards.
agreement at the margin. While uniform standards would imply a level of energy
The informational requirements that must be met efficiency that was ‘too high’ in low-price regions,
to identify an efficient tax regime, however, are reduced manufacturing costs would result in net cost
particularly onerous. The government must know not savings for most consumers.
only the level of consumer expectations but also the In terms of economic theory, government inter-
specific way in which they are formed, and this ventions are an appropriate means of reducing energy
information must be effectively conveyed to manu- intensity in markets where informational anomalies
facturers through the structure of the tax. In practice, inhibit the uptake of cost-effective energy-efficient
such information may be very difficult to obtain, technologies. Whether such interventions yield net
reducing the efficacy of tax instruments. reductions in energy use, however, depends on the
Such limitations suggest a potential role for the characteristics of the demand for energy services.
direct regulation of equipment performance. Energy Generally speaking, policy interventions will reduce
efficiency standards are a key element of energy policy not only energy intensity but also the cost of using
in the USA, where the Corporate Average Fuel energy services. Thus policies designed to improve
Economy standards led to demonstrable improvements energy efficiency may increase the level of energy-using
in the fuel economy of automobiles in the 1970s and activities ~ to so-called ‘take-back’ or ‘rebound’ effect
early 1980s (Greene [7]). State and local government (Greene [S]).
set requirements concerning the thermal performance In the context of the dynamic model, suppose that
of building elements, although such requirements are x=x(p+qe) with x’(.)<O is the demand for devices
often minimum acceptable standards which are when expected and realized energy intensity are
surpassed in practice by private contractors. stationary at the level e. The initial level of energy
If performance standards are to pass the test of intensity is e, =e: >eeS/ where ealS is the socially
economic efficiency, they must be set to equate the efficient level. Now assume that. the government
marginal costs and benefits of reducing the level of introduces performance standards that reduce energy
energy intensity. A good example of this approach is intensity to the efficient level and that expectations
embodied in the US appliance efficiency standards, converge to eeff over the long run. What is the
which are based explicitly on lifecycle cost criteria. resulting effect on energy use? We know that the price

270 ENERGY ECONOMICS October 1993

 Market harriers to energy eficiency: R. B. Howarth and B. Anderson

of devices is set equal to unit cost so that the stationary improvements in the fuel economy of light-duty
level of energy use is vehicles in the USA between 1966 and 1989 by only
5515%.
U(e)=e.x(p+qe)=e.x[c(e)+qe] (18)

A marginal decrease in e will reduce long-run energy Conclusion

use if the derivative of this function obeys the
restriction This paper has shown how the structural characteristics
of markets for energy-using equipment may impede
U’(e)=.u+e.[c’(e)+q]x’(.)>O the adoption of cost-effective energy-efficient tech-
(19)
nologies even when markets are competitive and all
market participants are rational. By rationality we
Algebraic manipulation yields the equivalent condition mean that agents do the best they can given constraints
on the resources and information at their disposal,
including limits on their cognitive capabilities. A
; [q+c’(e)] < -g$ (20)
central theme is that the costs incurred by individuals
in forming expectations of equipment performance
where P=p+qe is the total cost of device ownership. lead them to render decisions under incomplete or
Reduced energy intensity thus implies reduced imprecise information, biasing the market against
long-term energy use provided that the product of the investments in energy efficiency that would be made
energy intensity/ownership cost ratio times the wedge under perfect information.
between the price of energy and the marginal cost of While our arguments are theoretical and can be
reducing energy intensity is less than the negative corroborated only through careful empirical research,
inverse of the total cost elasticity of equipment demand we believe that they provide a degree of suport to
(the right-hand side of Equation (20)). Note that this Grubb’s [9] view of proper energy policy in a world
relation will hold unless the negative inverse elasticity of demand-side market barriers to energy efficiency
exceeds the share of energy as a fraction of total (Figure 1). In such a world, proper pricing and
ownership costs. As the energy cost share grows small government programmes designed to reduce market
the magnitude of the elasticity must become larger barriers to cost-effective energy efficient technologies
and larger for the relation to be violated. With an are complementary policy instruments. Only when the
energy cost share of lo%, for example, the elasticity two are implemented together is a socially efficient
must be - 10 or less if energy use is to grow with a resource allocation achieved.
reduction in energy intensity. Economists commonly argue that decentralized
In reality, energy costs are usually a small fraction private decision-making is preferable to collective
of the total cost of owning and operating energy-using action because it empowers individuals to make
equipment. Even in nations where gasoline is heavily choices that foster their personal well-being. While the
taxed, for example, fuel expenditures generally rationale for this perspective is compelling in many
account for less than 20% of the cost of owning and respects, we believe that enthusiasm for the market
using an automobile. In most cases, there is little should be attenuated by a realistic understanding of
evidence that equipment demand is unusually elastic the shortcomings and limitations of the market
with respect to ownership costs. Hence we should mechanism.
generally expect energy use to fall with policy-driven An individual acting alone is unlikely to hire an
reductions in energy intensity, although the decline in engineer to evaluate the costs and benefits of design
energy use may be less substantial than the reduction improvements intended to reduce the energy intensity
in energy intensity. of a new refrigerator or heating system. Minimum
This expectation is generally confirmed by empirical performance standards based on cost-effectiveness
studies of the effects of reduced energy intensities. criteria, on the other hand, might impose programme
Dubin et al [4], for example, compared engineering costs that are small in comparison with the associated
calculations of the energy savings achievable through benefits. Poorly designed regulations would deny
improvements in residential heating and cooling consumers access to desired attributes of energy-using
systems against actual performance. The engineering technologies. Well designed standards, in contrast,
calculations overestimated savings by l-13% for would limit consumers’ choices only in the trivial sense
cooling systems and 8812% for heating systems. of eliminating options that no one would willingly
Similarly, Greene [S] found that increased vehicle exercise given perfect information and unlimited
utilization reduced the energy savings associated with cognitive capacity.

ENERGY ECONOMICS October 1993 271

 Market barriers to energy e&ciency: K. B. Howarth and B. Andersson

References manufacturers, electric utilities, and the environment’,

Proceedings of the ACEEE Summer Study on Energy
1 K. Arrow, ‘Economic welfare and the allocation of EfJiency in Buildings, Vol 7, 1990, pp 107- 116.
resources for inventions’, in R. Nelson, ed, The Rate 16 D. L. McNeil1 and W. L. Wilkie, ‘Public policy and
and Direction of Inventive Activity, Princeton University consumer information: impact of the new energy labels’,
Press, Princeton, NJ, 1962. Journal of Consumer Research, Vol 6, 1979, pp 1~ 11.
2 R. Carlsmith, W. Chandler, J. McMahon and D. 17 S. M. Nadel and K. M. Keating, ‘Engineering estimates
Santino, Energy EfJiciency: How Far Can We Go?, Oak vs. impact evaluation results: how do they differ and
Ridge National Laboratory ORNL/TM-11441, 1990. why?‘, paper presented to the Conference on Energy
3 C. Difiglio, K. G. Duleep and D. L. Greene, ‘Cost Program Evaluation: Uses, Methods, and Results,
effectiveness of future fuel economy improvements’, 1991.
Energy Journal, Vol 11, No 1, 1990, pp 65-86. 18 J. B. Robinson, ‘The prooiof the pudding ~ making
4 J. Dubin, A. Miedema and R. Chandran, Price Effects energy efficiency work’, Energy Policy, Vol 7, 1991,
of Energy-EfJient Technologies.. A Study of Residential pp 631-645.
Demand for Heating and Cooling, Research Triangle 19 A. Rosenfeld, The Role of Federal Research and
Institute, 1982. Development in Advancing Energy EfJiency. Con-
5 A. Fisher and M. Rothkopf, ‘Market failure and energy gressional testimony on the Department of Energy
policy’, Energy Policy, Vol 17, No 4,1989, pp 3977406. Conservation Budget Request, Lawrence Berkeley
6 H. Geller, J. Harris, M. Ledbetter, M. Levine, E. Mills, Laboratory, 1991.
R. Morris and A. Rosenfeld, ‘The importance of 20 H. Ruderman, M. D. Levine and J. E. McMahon, ‘The
government-supported research and development in behavior of the market for energy efficiency in
advancing energy efficiency in the Unites States building residential appliances including heating and cooling
sector’, in T. Johansson, B. Bodlund and R. Williams, equipment’, Energy Journal, Vol 8, No 1, 1987,
eds, Electricity - Efficiency End-Use and New pp 101-124.
Generation Technologies, and Their Planning Implica- 21 T. Russell and R. Thaler, ‘The relevance of quasi
tions, Lund University Press, Lund, 1989. rationality in competitive markets’, American Economic
7 D. L. Greene, ‘CAFE or price? An analysis of the effects Review, Vol 75, 1985, pp 1071-1082.
of federal fuel economy regulations and gasoline price 22 A. H. Sanstad, J. G. Koomey and M. D. Levine,
on new car MPG, 1978- 1989’, Energy Journal, Vol 11, On the Economic Analysis of Problems in Energy
No 3, 1990, pp 37-57. Eflciency: Market Barriers, Market Fuilures, and
8 D. L. Greene, ‘Vehicle use and fuel economy: how big Policy Implications, Lawrence Berkeley Laboratory, in
is the “rebound” effect?‘, Energy Journal, Vol 13, No 1, preparation.
1992, pp 1177143. 23 L. Schipper, Raising the productivity of energy
9 M. Grubb, Energy Policies and the Greenhouse Eflect. utilization’, Annual Review of Energy, Vol 1, 1976,
Vol 1: Policy Appraisal, Dartmouth, Royal Institute of pp 4555517.
International Affairs, 1990. 24 P. C. Stern, ‘Blind spots in policy analysis: what
10 J. A. Hausman, ‘Individual discount rates and the economics doesn’t say about energy use’, Journal of
purchase and utilization of energy-using durables’, Bell Policy Analysis and Management, Vol 5, 1986,
Journal of Economics, Vol 10, 1979, pp 33-54. pp 200-227.
11 M. I. Kamien and N. L. Schwartz, Market Structure 25 R. J. Sutherland, ‘Market barriers to energy-efficiency
and Innovation, Cambridge, Cambridge University investments’, Energy Journal, Vol 12,1991,pp 15-34.
Press, 1982. 26 R. D. Tollison and R. B. Ekelund Jr, ‘The economic
12 F. Krause and J. Eto, Least-Cost Utility Planning. A efficiency of federal versus state energy conservation
Handbook for Public Utility Commissioners. Vol 2: standards for appliances’, unpublished manuscript, 1987.
The Demand Side: Conceptual and Methodological 27 US Department of Energy, Technical Support Document:
Issues, Lawrence Berkeley Laboratory, 1988. Energy Conservation Standards.for Consumer Products
13 M. Ledbetter and M. Ross, ‘A supply curve of conserved - Dishwashers, Clothes Washers, and Clothes Dryers,
energy for automobiles’, paper presented to the DOE/CE-0229p, 1990.
Intersociety Energy Conversion Engineering Conference, 28 US National Academy of Sciences, Policy Implications
New York, NY, 1990. of Greenhouse Warming ~ Report of the Mitigation
14 A. B. Lovins and L. H. Lovins, ‘Least-cost climatic Panel, Washington, DC, National Academy Press,
stabilization’, Annual Review of Energy and the 1991.
Environment, Vol 16, 1991, pp 433-531. 29 D. Wilson, L. Schipper, S. Tyler and S. Bartlett, Policies
15 J. E. McMahon, D. Berman, P. Chan, T. Chan, J. and Programs for Promoting Energy Conservation in the
Koomey, M. D. Levine and S. Stoft, ‘Impacts of US Residential Sector: Lessonsfrom Five OECD Countries,
appliance energy performance standards on consumers, Lawrence Berkeley Laboratory LBL-27289, 1989.

272 ENERGY ECONOMICS October 1993